Is it possible that I am requesting an EC2 instance but cannot execute it?
Edit: This question should NOT ask about "Spot Instances"; this question is to ask the regular "On Demand Instance". I think I need to clarify this by reading the comments below.
Basically, my question is whether I should take into account the risk that when I need to start an EC2 instance, but that the EC2 scope has finished working and cannot fulfill my request.
I understand that the chances of the above situation are extremely low, but I would like to understand if AWS has any SLA to ensure that the situation / risk does not happen.
source to share
Security controls exist to make it unlikely that a specific type of instance in a specific Availability Zone will be unavailable at a specific time ... but it is possible, and there are no guarantees provided by AWS that a given type of EC2 instance will be available to run on demand at any given time. in any specific Availability Zone , unless you purchased Reserved Instances of that type, especially in that Availability Zone. This assumes that there is always enough hardware available so that you can at least have a number of paid Reserved Instances, including the ability to launch enough new instances to bring the total to that minimum.
Reserved Instances are usually discussed in the context of their associated discount, but they have two purposes:
Reserved Instances are not physical instances, but rather a billing discount applied to use the On-Demand instances in your account. These on-demand instances must meet certain attributes in order to benefit from the billing discount.
When you purchase Reserved Instances in a specific Availability Zone, they provide capacity reservation. (emphasis mine)
For example, if you purchased 4 t2.2xlarge Reserved Instances in us-east-2a, be sure that you can always run enough to bring the total running instances of this type in this zone to 4. If you already have 4 , not sure if you can start more, but there is confidence that if you stop them, you can start them again.
The pricing models for Reserved Instances have changed over the years, so Reserved Instances tend to be billed at the same rate whether they work or not, so you can look at it in one of two ways:
If you need capacity all the time, you get a substantial discount ... or if you don't need constant capacity, you pay all the time for equipment that is largely unnecessary, but you still pay less than you pay for on-demand copies. without reservations, and you can either leave it running or run it whenever you need it.
Should we consider the risk that the entire region has capacity problems? You have to consider this, but there are, historically speaking, other significant outage scenarios that are more likely ... EBS and S3 had failures that disrupted the ability to launch instances, even though bandwidth was idle in EC2.
source to share
Yes . It may be that your instance request could not be fulfilled. An Instance on Demand does not guarantee you an Instance. Specifically, the instances t2.small
will most likely fail based on my experience. AWS may have a limited number of instances t2.small
.
How can you make sure it is always executed?
reserve copies for you so that it is not shared with anyone else. But there are costs associated with this. You pay for a copy, regardless of whether you use it. I am speaking in general terms. Reserved Instance is a tricky topic, but it is a route you must take if you want AWS to guarantee you an instance.
source to share
Answer: Yes, your launch request might fail because there is no available capacity in the corresponding AZ. I would say that this is a rare occurrence, but certainly possible.
You can mitigate by using multiple zones in the same region, other regions, or using Reserved Instances .
source to share
This happened today on an AWS account that I manage. If I am not mistaken it was the r4 family, exactly an instance of r4.xlarge (4 hours ago) in the Virginia region. I had to choose a different A-Z. This is why AWS always recommends working with multiple AZs.
But for this reason I started using Scheduled Reserved Instances
Over a period of time, you will always have a Family Instance Type Reserved Capacity.
Of course, it works if you have a specific schedule or workflow.
Hope this helps you.
source to share